By using the laser-induced breakdown spectroscopy (LIBS) technique, the plasma emission spectra of the series of geological standard samples from United States Geological Survey (USGS) are obtained. By using the artificial neural network, the content of Fe in different USGS geological standard samples is measured. The relative error between the measured content and standard content of BCR-1G, BHVO-2G, BIR-1G, GSD-1G, and GSE-1G is 1.86%, 5.73%, 0.27%, 3.86% and 2.63%, respectively, which shows that LIBS combined with artificial neural network can measure the Fe content well in the series of geological standard samples from USGS.